How Do I Track Thee, Mobile Shopper? Let Me Count The WaysWritten by Evan Schuman
For quite a few years now, retailers have salivated over the idea of mobile phones revealing exactly where shoppers are at all times. Retailers would know which displays customers are standing in front of, for how long and what actions they take right afterward. Unfortunately, even though mobile devices have advanced quite a bit recently, the ability to know location with any precision has been elusive.
No major advances in mobile location technology have emerged. In the last couple of months, however, quite a few very different approaches to location tracking have emerged. These range from leveraging the earth’s magnetic field to piggybacking the data already used by mobile ads, tracking a combination of Wi-Fi and Bluetooth signals, and riding the audio signals from existing music speakers. One app even reacts to light patterns from specially enhanced LED bulbs.
But none of these approaches, thus far, has proven effective at delivering truly precise locations. For example, a customer is looking for a very specific cereal, such as multigrain oats and honey flavored Special K. With item-level product tracking, the system could go far beyond directing the shopper to the cereal aisle. It would offer mobile app-based directions right to the Special K area and then navigate (up one shelf, three inches to the right, third row) the shopper to the exact SKU.
It’s a given that this nirvana app has already checked real-time inventory to make sure the specific product really is in stock and then used video or RFID tracking to verify the product is indeed where it’s supposed to be. (Come to think of it, a truly impressive demo would be if this app directed the shopper to where the one box of that flavor of cereal had been misplaced by another shopper.)
Now that the resistance to using such customer-tracking approaches appears to be slightly melting (well, maybe not), it’s something that IT shops are going to need to examine more thoroughly.
Alas, location services don’t come close to that level of capability yet, under any of these approaches. But it should happen soon. (Recent trials by Google and a more limited exploration by grocery giant Meijer have amply demonstrated that we’re far from the goal.)
To get there, of course, will require major improvements in tracking all three of the elements needed: the shopper, the desired product and the inventory data, so a non-existent product can be immediately flagged—ideally with a message saying, “I’m so sorry but we’re out of stock at the moment. Here are five suggestions of similar products that we do have in stock. Here’s a button to special order the original product right now, and we’ll E-mail or message you when it’s in. Or click on this button and we’ll ship it to you through our E-Commerce operation.”
Beyond the upsell potential—which Walmart is working hard to use for its in-aisle mobile checkout trial—there is, of course, the huge CRM potential of knowing everything a customer is even thinking about buying (through a barcode scan) and then marrying that information to the shopper’s location.
Let’s look at some of the location options:
A Finland-based vendor called IndoorAtlas said that the nature of many retail buildings makes them natural for magnetic mapping.
“Steel masses inside buildings twist Mother Earth’s magnetic field such that every spot produces a unique pattern. Each building, floor and corridor creates a distinct magnetic field disturbance that can be measured to identify a location and generate a map,” the vendor says on its site.
This approach currently delivers roughly one-meter accuracy (half-meter in either direction) with current handsets, said IndoorAtlas CEO Janne Haverinen. But he said with sensor improvements expected within the next two to three years, he believes accuracy will improve 10 times—to about 10 centimeters (or approximately 4 inches).
To set it up initially requires a detailed mapping of the store, something that Haverinen said typically takes about two-and-one-half hours for a 10,000-square meter (or about 110,000 square feet) store. Given the magnetic nature of it, retailers considering such an approach need to be aware of anything metal (specifically, anything that sticks to a magnet). When any such metal is changed, the store—or at least the portion of the store where the change has happened—needs to be remapped. “So if you remove or rearrange metal shelves, it will change the magnetic field,” Haverinen said.
Metal shopping carts pose a smaller issue. “If someone passes you with a metal cart, it might have a small effect. It might give you some bad readings,” he said, adding that the software could—theoretically—filter that out. Either way, Haverinen said, the effect would be very short-lived.
His pricing approach, though, could be more disruptive. Instead of a flat rate, IndoorAtlas wants to charge retailers for every time any shopper in the store makes a location interaction. Presumably, it could also be triggered in reverse, where the store activates a “where is the customer now?” inquiry. For a high-volume store, the CEO said the per interaction cost would be “a small fraction of a cent” and then, when asked, said “one-millionth of a penny is close.”
The common thread in almost all these approaches is to work with something that already exists in the mobile or retail environment. What a few vendors are doing is trying to piggyback location information on top of the existing back-and-forth signals from the networks that decide which mobile ads to post. Those data exchanges already identify the shopper in various ways, including the number associated with specific hardware (that shopper’s phone), the specific application, the phone’s OS and other attributes. Those small snippets of ad-serving codes can be picked up by anyone with access to that ad network, said Sense Networks CEO David Petersen. “These are true IDs, analogous to cookies. When we see that, it bounces that location against local points of interests” and looks for a match, he said.
The problem is that this system is almost exclusively designed for ad transmission, so the location precision is not going to be especially precise. But retailers building atop such a system could work around that issue.
Although not related to location, our personal favorite example of leveraging mobile data to identify the phone came from a Seattle vendor in February. It avoided a PIN by grabbing whatever data about the phone it could, including the list of installed apps and the names of the five most frequently called friends. Other attributes grabbed include the operating system version, an app cookie, the SD card, the nature of a Wi-Fi connection, and carrier and CPU performance.